Deep learning has been successfully applied for predicting asset prices using financial time series data. However, image-based deep learning models excel at extracting spatial information from images and their potential in financial applications has not been fully explored. Here we propose a new model---channel and spatial attention convolutional neural network (CS-ACNN)---for price trend prediction that takes arbitrary images constructed from financial time series data as input. The model incorporates attention mechanisms between convolutional layers to focus on specific areas of each image that are the most relevant for price trends. CS-ACNN outperforms benchmarks on exchange-traded funds (ETF) data in terms of both model classification metrics and investment profitability, achieving out-of-sample Sharpe ratios ranging from 1.57 to 3.03 after accounting for transaction costs. In addition, we confirm that the images constructed based on our methodology lead to better performance when compared to models based on traditional time series data. Finally, the model learns visual patterns that are consistent with traditional technical analysis, providing an economic rationale for learned patterns and allowing investors to interpret the model.
We propose several methods to obtain endogenous and positive ultimate forward rates (UFRs) for risk-free interest rate curves based on the Smith-Wilson method. The Smith-Wilson method, adopted by Solvency II, can both interpolate the market price data and extrapolate to the UFR. However, it requires an exogenously-chosen UFR. de Kort and Vellekoop (2016) proposed an optimization problem to obtain an endogenous UFR. In this paper, we prove the existence of the optimal endogenous UFR to their optimization problem. In addition, in order to ensure the positiveness of the optimal UFR, we formulate a new optimization framework with nonnegative constraints. Furthermore, we also propose another optimization framework to generate endogenous and positive UFRs with prior knowledge. The feasibilities of both methods are proven under several mild conditions. We use Chinese government bond data to illustrate the capabilities of our methods and find the dynamic behaviour of Chinese risk-free interest rate curves.
Semi-competing risks refer to the phenomenon where a primary outcome event (such as mortality) can truncate an intermediate event (such as relapse of a disease), but not vice versa. Under the multi-state model, the primary event is decomposed to a direct outcome event and an indirect outcome event through intermediate events. Within this framework, we show that the total treatment effect on the cumulative incidence of the primary event can be decomposed into three separable pathway effects, corresponding to treatment effects on population-level transition rates between states. We next propose estimators for the counterfactual cumulative incidences of the primary event under hypothetical treatments by generalized Nelson-Aalen estimators with inverse probability weighting, and then derive the consistency and asymptotic normality of these estimators. Finally, we propose hypothesis testing procedures on these separable pathway effects based on logrank statistics. We have conducted extensive simulation studies to demonstrate the validity and superior performance of our new method compared with existing methods. As an illustration of its potential usefulness, the proposed method is applied to compare effects of different allogeneic stem cell transplantation types on overall survival after transplantation.